Enhanced Methods of Denoising and Segmenting MRI Images for Brain Tumor Detection

Abstract

Image processing is used to improve the quality of the image for better human understanding. MRI images are best method for identifying the abnormal tissue growth in brain whereas CT scan is best for bone related problems. Soft tissues related problems will be shown best in MRI scan. Quality of the image is not maintained because of noise. Sometimes there is a possibility that because of noise the physician may diagnose MRI image wrongly. In order to avoid such kind of wrong diagnosis, machine learning techniques like Support Vector Machine are suggested. In this work, various denoising models, segmentation model and classification technique is proposed. In the first work, different types of noise are removed by a novel model for denoising of two dimensional, three dimensional images based on Discrete Wavelet Transformation and Dynamic Thresholding was proposed. In the first work, different types of noise are removed by a novel model for denoising of two dimensional, three dimensional images based on Discrete Wavelet Transformation and Dynamic Thresholding was proposed. This model proposes different types of thresholding techniques. A quantization scheme that improves the compression ratio and quality of the reconstructed image was proposed. Several parameters are evaluated like PSNR, MSE are calculated and it seems PSNR values are comparatively high than the existing method. It also proves that, MSE values are very less which means lower the error rate and higher the clarity of image. newline